am I low on GFlops?

Associate
Joined
28 Oct 2012
Posts
101
Hi,

I recently overclocked my PC and I'm having very stable temperatures around 55 degrees when running Intel Burn Test.

The test however takes ages and when it finishes a loop it's usually around 33 Gflops, not entirely sure what this mean or to what extent it'll affect my performance but I've seen people run the test in less than half my time and reach more than double the Gflop

my CPU is an i5 2.67Ghz overclocked to 3.6 GHz

I can provide a screenshot later today once I'm home again but just curious if this is normal and how it affects my performance.
 
Last edited:
Which IBT version are you using and what cpu specifically? (im going to assume an first gen i5 750 from the stock clockspeed) If you are on a first gen, IBT will be running without AVX instruction set which is what gives the doubling of flops with sandy/ivy over the first gen i processors.

However having said that that you ought to be getting around/near 50 Gflops at 3.6 (with all programs closed, AV etc as it does affect the output)

what IBT settings did you use and is the output variable? Could indicate throttling/instability if the numbers jump about/dip randomly.

Ram can also impact IBT decently so the speed and timing can also play a factor here.
 
Last edited:
iirc u need to run custom ram amount, as if u are using cached ram, it can drop the gflops.
open task manager, then performance, note how much free ram is stated.
in IBT, enter just below this in the custom amount, eg. 2560mb free, then enter 2500mb, then 4 threads and run again.
have antivirus, web browsers etc.. closed as this can affect the speed.
ideally u are looking for the ram to run at at least 75%, to get a proper reading.

more details in this thread.
http://forums.overclockers.co.uk/showthread.php?t=18297411
 
I'll try a few more things as soon as I get home in about 2 hours. My CPU is a:
Intel Core i5 750 2.66GHz Socket LGA1156 8MB L3 Cache Retail Boxed Processor

I have 16 GB of Ram and I was using 12000 for the test. I'll give you guys more info once I can get it. Thank you so much!
 
ok, I'm finally home and this is the result I got. Only ran it ocne since it takes so long:

Time (s)
1544.068

Speed (GFlops)
33.1233

Result
3.601843e-002


1544 seconds is a tad bit longer than what I expected but I think the result is correct at least. The test was run on all threads

thPeoLF.png
2Hzxte1.png
 
Last edited:
Can you try 10-15 runs at 1024 and 2048 MB. At 12000 becasue each run takes so much time any throttling/instability unlikely show from one run to the next, and the variances merge and give a consistent result.

Also after the above, try lowering the cpu oc (maintain voltages) and see if it remains the same/improves the 12000MB run result, Or alternatively do that comparison against the above 1024/2048 runs.
 
Last edited:
did a 2048 run now and these were the results:
Qw26V1c.png


looking a lot better, I'm considering going for the 4 GHz if I can get everything fully stable and secure seeing as I'm only on 1.2V yet and my temperature doesn't even reach 60 degrees
 
Yup that looks like all ok, sounds like j.col had it spot on with the ram setting going into cached ram.

Go for it, above 3.8 the first gen start really eating volts. Happy ocing!
 
i have the same cpu at 3.8ghz, and i get around 55 gflops, so its within reason.
it is important when running IBT to use free ram only, as this will improve the gflops score.
and as i said, u need to be around 75% ram usage as well if possible.
have a look in the thread i linked above, then read wingzero's and joeyjo's posts, these will actually explain the process and the actual amount of gflops u can expect.
 
Back
Top Bottom